CN115698880A

CN115698880A - Abnormal modulation cause display device, abnormal modulation cause display method, and abnormal modulation cause display program

Info

Publication number: CN115698880A
Application number: CN202180039068.9A
Authority: CN
Inventors: 小园英俊; 武次祐树; 近藤弘康
Original assignee: Daicel Corp
Current assignee: Daicel Corp
Priority date: 2020-05-29
Filing date: 2021-05-25
Publication date: 2023-02-03
Also published as: EP4160338A1; US20230205194A1; WO2021241577A1; JPWO2021241577A1

Abstract

The accuracy of determination of the cause of abnormal modulation in a production apparatus is improved, and an appropriate countermeasure is proposed to the user. The abnormal modulation cause display device is provided with: a process data acquisition unit that reads process data from a storage device that stores process data continuously output from a plurality of sensors provided in a production facility; an abnormality determination unit that calculates an abnormality degree indicating a degree of modulation of the process data read by the process data acquisition unit; a cause diagnosing unit that determines whether or not the degree of abnormality calculated by the abnormality determining unit satisfies a predetermined criterion with respect to the process data output from the plurality of sensors, using cause and effect relationship information that defines a combination of a cause and a modulation of the process data output from the plurality of sensors that appears as an influence due to the cause; and an output control unit that reads out information indicating a countermeasure to be taken for the reason from a storage device that also stores the information indicating the countermeasure, and outputs the information to an output device.

Description

Abnormal modulation cause display device, abnormal modulation cause display method, and abnormal modulation cause display program

Technical Field

The present disclosure relates to an abnormal modulation cause display device, an abnormal modulation cause display method, and an abnormal modulation cause display program.

Background

Conventionally, there has been proposed a plant control monitoring apparatus including a display device for representing an operation state of each of a plurality of devices, machines, and processes included in a plant by representing a process state signal output from the devices, machines, and processes by using a predetermined mark, the display device including at least a monitoring screen for monitoring an operation of the plant, a guidance screen for responding to the alarm, the guidance screen including guidance selected according to an importance level of the alarm, and an operation screen for operating the plant, the display device being configured to display, on the display screen displayed on the display device, a guidance screen indicating whether or not the alarm corresponds to the alarm when the alarm occurs, and the plant control monitoring apparatus further including a first input unit for calling the guidance screen corresponding to the alarm on the display screen when the alarm occurs (patent document 1).

Documents of the prior art

Patent document

Patent document 1: japanese patent No. 3699676

Disclosure of Invention

Problems to be solved by the invention

In general, in a production facility, it is desirable to prevent abnormal preparation and to suppress influences on safety, stability, quality of a product, cost, and the like. The present technology aims to improve the accuracy of determining the cause of abnormal modulation in a production facility and to propose appropriate countermeasures to a user.

Technical scheme

The abnormal modulation cause display device is provided with: a process data acquisition unit which reads out process data from a storage device which stores process data continuously output from a plurality of sensors provided in a production facility; an abnormality determination unit that calculates an abnormality degree indicating a degree of modulation of the process data read by the process data acquisition unit; a cause diagnosing unit configured to determine whether or not the degree of abnormality calculated by the abnormality determining unit satisfies a predetermined criterion with respect to the process data output from the plurality of sensors, using cause and effect information defining a combination of a cause and a modulation of the process data output from the plurality of sensors, the modulation occurring as a result of an influence of the cause; and an output control unit that reads out information indicating a countermeasure to be taken for the reason from a storage device that also stores the information indicating the countermeasure, and outputs the information to an output device.

By using the causal relationship information as described above, among the process data output from the plurality of sensors affected by the cause, the process data whose degree of abnormality satisfies a predetermined criterion increases, and thus, it is possible to sense an abnormality due to some cause. In other words, even if the process data output from the plurality of sensors has a small number of process data whose degree of abnormality satisfies a predetermined criterion, the process data can sense a sign of abnormality caused by some cause. Therefore, the performance of determining the cause of abnormal modulation in the production apparatus can be improved. Further, based on the cause determined in this manner, an appropriate countermeasure according to the situation can be presented to the user.

The cause diagnosing unit may multiply the process data of a plurality of types by a coefficient corresponding to the type of the process data or a coefficient based on the magnitude of the abnormality degree to obtain the accuracy of the cause of the modulation, and the abnormal modulation cause display device may further include an output control unit that outputs, to the output device, a plurality of candidate causes that can cause the modulation and the accuracy of the cause, for the modulation. Based on the reason for the candidate and its accuracy, the user can select an appropriate countermeasure.

Further, the output control unit may output, to the output device, a logical tree in which the modulation is the root, the cause of the modulation is the leaf, and the event connection components that occur during the period from the cause to the modulation are layered, based on the causal relationship information, and may output, to the output device, information indicating a measure to be taken for the cause in association with the cause. According to such a logical tree, events occurring in the process of modulation and causes of modulation can be displayed in a form that is easily visible to the user.

It should be noted that the contents described in the claims may be combined as much as possible without departing from the scope of the problem and technical idea of the present disclosure. Further, the contents of the technical solution can be provided as a system including a device or a plurality of devices such as a computer, a method executed by a computer, or a program causing a computer to execute. A recording medium for storing the program may be provided.

Effects of the invention

According to the disclosed technology, it is possible to improve the accuracy of determining the cause of abnormal modulation in a production apparatus, and to propose appropriate countermeasures to a user.

Drawings

Fig. 1 is a diagram showing an example of a system according to the present embodiment.

Fig. 2 is a schematic diagram showing an example of a process performed by a machine provided in a plant.

Fig. 3 is a diagram for explaining an example of process data in a batch process.

Fig. 4 is a diagram showing an example of a preset process line definition table.

Fig. 5 is a diagram showing an example of a preset tag definition table.

Fig. 6 is a diagram for explaining an example of process data in the continuous process.

Fig. 7 is a diagram showing an example of the trace back information.

Fig. 8 is a diagram for explaining the correspondence between process data in the continuous process and production numbers in the batch process.

Fig. 9 is a diagram showing an example of information registered in advance in the knowledge base.

Fig. 10 is a diagram showing an example of a logical tree showing a relationship between modulation and its cause.

Fig. 11 is a diagram for explaining the synchronous processing of the process data.

Fig. 12 is a diagram for explaining an example of calculating the degree of abnormality from the distance from the reference with respect to time-series data.

Fig. 13 is a diagram for explaining an example of calculating the degree of abnormality from the distance from the reference in consideration of the positive and negative directions with respect to the time series data.

Fig. 14 is a diagram for explaining abnormality sensing using a self-encoder.

Fig. 15 is a block diagram showing an example of the configuration of the abnormal modulation cause specifying device.

Fig. 16 is a process flowchart showing an example of the learning process performed by the abnormal modulation cause specifying device.

Fig. 17 is a diagram showing an example of the action table.

Fig. 18 is a process flowchart showing an example of the abnormality sensing process performed by the abnormality modulation cause determination device.

Fig. 19 is a diagram showing an example of a screen to be output to the input/output device.

Fig. 20 is a diagram showing another example of a screen output to the input/output device.

Detailed Description

Hereinafter, an embodiment of the abnormal modulation cause specifying device will be described with reference to the drawings.

< embodiment >

Fig. 1 is a diagram showing an example of a system according to the present embodiment. The system 100 includes an abnormal modulation cause determination device 1, a control station 2, and an assembly 3. The System 100 is, for example, a Distributed Control System (DCS) and includes a plurality of Control stations 2. That is, the control system of the plant unit 3 is divided into a plurality of zones, and the control zones are controlled by the control station 2 in a distributed manner. The control station 2 is a conventional device in the DCS, and receives a status signal output from a sensor or the like provided in the plant unit 3 or outputs a control signal to the plant unit 3. Then, actuators such as valves and other devices provided in the unit 3 are controlled based on the control signals.

The abnormal modulation cause determination apparatus 1 acquires a status signal (process data) of the plant unit 3 via the control station 2. The process data includes the temperature, pressure, flow rate, etc. of the raw material and the intermediate product to be processed, and set values for determining the operating conditions of the equipment provided in the plant 3. Further, the abnormal modulation cause determination apparatus 1 creates an abnormality sensing model based on a knowledge base that stores the correspondence relationship of the assumed cause and the influence that occurs as an abnormality, for example. For example, the following model is created: based on a knowledge base for determining abnormal modulations, their precursors, their causes based on methods where the changes in the sensed process data deviate from an allowed range. Then, the abnormal modulation cause determination apparatus 1 can sense the occurrence of abnormal modulation or its precursor using the model and the process data. The abnormal modulation cause specification device 1 may determine, for example, a candidate operation condition for suppressing abnormal modulation based on a table storing causes of abnormal modulation and actions for coping with the causes and the specified causes, and present the operation condition to the user.

Fig. 2 is a schematic diagram showing an example of a process performed by a machine provided in a plant. In the present embodiment, the process may include a batch process 31 and a continuous process 32. In the batch step 31, the processing target is sequentially processed for each predetermined processing unit, for example, the processes of receiving, holding, and discharging the raw material of each apparatus are sequentially performed. In the continuous step 32, the continuously introduced processing objects are continuously processed, for example, the processes of receiving, holding, and discharging the raw materials are performed in parallel. Furthermore, the process may include a plurality of series 33 performing the same process in parallel.

The apparatuses for performing the respective treatments include, for example, a reactor, a distillation apparatus, a heat exchanger, a compressor, a pump, a storage (tank), and the like, and these apparatuses are connected via pipes. Further, sensors, valves, and the like are provided at predetermined positions of the devices and pipes. The sensors may include thermometers, flow meters, pressure gauges, level meters, concentration meters, and the like. Further, the sensor monitors the operation state of each device and outputs a state signal. The sensors provided in the unit 3 are sensors to which "tags" are added as identification information for identifying the respective sensors. That is, the kind of process data can be determined based on the tag. Then, the abnormal modulation cause determination device 1 and the control station 2 manage input/output signals to/from the respective machines based on the tags.

< batch Process >

Fig. 3 is a diagram for explaining an example of process data in a batch process. The left side of fig. 3 shows a part of the process of the batch process 31 shown in fig. 2. Specifically, the process includes a pulverizer 301, a cyclone 302, a pretreatment 303, a pre-cooler 304, and a reactor 305. These processes are classified into a pretreatment step, a precooling step, and a reaction step. The right column of fig. 3 shows an example of process data acquired in each process. In the preprocessing, time series data are acquired from sensors labeled 001 and 002. In the pre-cooling process, time series data is acquired from sensors labeled 003 and 004. In the reaction process, time series data is acquired from the sensors labeled 005, 006, and 007. In the batch process, a processing object associated with a production number (also referred to as a "product number", a "lot number", or a "management number") is intermittently processed. That is, the production number is identification information for identifying a processing object that is collectively processed in the batch process. As shown in fig. 3, data on the timing of the processing object that is subsequently associated with the production number is obtained as time passes. In the present embodiment, the control station 2 manages the production number and indicates the stage of processing in the process after the subdivision into the batch process. When the procedure is reset by a PLC (Programmable Logic Controller, sequencer) in the plant unit 3 connected to the control station 2, the production number of the process data output from the control station 2 may be used as appropriate (for example, after a set time has elapsed after the procedure is switched in the PLC) according to the timing of communication between the control station 2 and the plant unit 3. The setting time may be set for each manufacturing line and for each step after the division.

Fig. 4 is a diagram showing an example of a preset process line definition table. In the process line definition table, production numbers, definitions of steps indicating the stages of processing in each process, and the types of processing objects to be processed in each process are registered for each series and process. The process line definition table may be a so-called database table or a file of a predetermined format such as CSV. The process line definition table is also created by the user in advance and read out by the abnormal modulation factor specifying device 1.

The process line definition table includes attributes of series, process, product number, step, and variety. Identification information for identifying a series of processes is registered in a column of the series. Identification information indicating a process after division in a batch process is registered in a process column. A production number as identification information for identifying a processing object collectively processed in the batch process is registered in a column of the product number. Definitions of timings of a plurality of steps indicating stages of processing in the process are registered in the step column. The category of the processing target is registered in the column of the item.

Fig. 5 is a diagram showing an example of a preset tag definition table. The tag definition table defines the acquisition timing of the process data obtained from the sensor corresponding to each tag.

The tag definition table may be a table of a so-called database, or may be a file of a predetermined format such as CSV (Comma Separated Values). The tag definition table is created by the user in advance and read out by the abnormal modulation cause determination device 1.

The label definition table includes attributes of labels, series, processes, and collection intervals. A tag that is identification information of the sensor is registered in a column of the tag. Identification information for identifying a series of processes is registered in a column of the series. Identification information indicating a process after division in a batch process is registered in a process column. Information indicating the interval at which the output value of the sensor is acquired is registered in the column of the collection interval.

< continuous Process >

Fig. 6 is a diagram for explaining an example of process data in the continuous process. The left side of fig. 6 shows a part of the process of the continuous process 32 shown in fig. 2. Specifically, the process includes a reservoir 311 and a pump 312. The right column of fig. 6 shows an example of process data acquired in each process. In the continuous process 32, time-series data corresponding to the label but not corresponding to the production number is continuously acquired from the sensor. In the continuous process, time series data is acquired from each sensor labeled 102 and 103. In the continuous process, the machine continuously receives the processing object and continuously performs the processing.

In the case where the continuous process is performed after the batch process, the traceability information set by the user in advance is used in the present embodiment in order to associate the processing object in the batch process with the processing object in the continuous process. Fig. 7 is a diagram showing an example of the trace back information. The trace back information includes attributes of sampling interval and dwell time. In the column of the sampling interval, an interval at which sampling for process inspection is performed in a continuous process, for example, by a division method, is registered. The column of the residence time registers the time that the processing target stays from the completion of the batch process until the process included in the continuous process is reached.

Fig. 8 is a diagram for explaining the correspondence between process data in the continuous process and production numbers in the batch process. The process data is acquired at intervals set in the trace back information, for example. In the case where the continuous step is performed after the batch step, a product obtained from the batch step completed in a predetermined period is introduced into a storage or the like as a processing target of the continuous step. Therefore, the process data in the continuous process can be associated with the completion time of the batch process and the production number group included in the predetermined period by tracing back the residence time of the processing target from the completion of the batch process to the measurement of the sensor. By establishing the correlation in this manner, it is possible to improve the accuracy of determining the cause of an abnormality using process data in the batch process when the batch process and the continuous process are continuously performed.

As described above, by associating the production number in the batch processing with the measurement timing in the continuous process, the accuracy of determining the cause of the abnormality can be improved.

Fig. 9 is a diagram showing an example of information registered in advance in the knowledge base. The knowledge base is stored in advance in the storage device of the abnormal modulation cause specification device 1. The table of fig. 9 includes columns of "influence" corresponding to the respective sensors (tags) and rows of "assumed cause" indicating modulation. That is, the direction of the fluctuation of the value is registered in the column corresponding to the sensor affected by the causes such as "cause 1" and "cause 2" indicated in each row. In the knowledge base, "up" indicating an increase (rise) in the output value of the sensor or "down" indicating a decrease (fall) in the output value of the sensor indicates the direction of the change.

As shown in fig. 9, the combination of the cause and the influence is not limited to one-to-one. Further, a calculation method of process data, extraction timing, a threshold value for abnormality determination, and the like are determined in association with each sensor. Information indicating the calculation performed on the output value of each sensor is registered in the row of the calculation method. In the present embodiment, for example, the calculation is performed by using a machine learning method such as the hotelling method, the k-nearest neighbor algorithm, the DTW center Averaging (DTW Barycenter Averaging), the auto encoder (auto encoder), and the graph Lasso (Graphical Lasso). In the row of the extraction timing, information indicating the timing at which the value used for abnormality determination is extracted from the output values of the sensors is registered. The timing may be defined, for example, in the batch process, by a step indicating a stage of the process in each step, a specific period, a time point, or the like. Note that the continuous processing may be defined by a sampling interval as shown in fig. 7. In the row of the threshold values, threshold values are registered as references for determining abnormality in each abnormality determination method. The threshold value includes, for example, both an upper limit and a lower limit. As described above, the knowledge base defines a combination of causal events and causal relationships that are the effects of modulation of process data generated thereby. The combination of causal relationships can be expressed in a tree form in which events occurring in the process from the cause to the modulation due to the factor She Laijiang are assumed to be connected in a hierarchical manner along the time sequence, with the modulation occurring as the root.

The knowledge base is set up to be created in advance by the user, for example, based on the HAZOP (Hazard and Operability Study: risk and Operability analysis). HAZOP is used for example for the following methods: the monitoring range is a set point of an alarm, a management range (a threshold value of an upper limit and a lower limit), a deviation from the management range (abnormality, modulation), an enumeration of assumed causes of deviation from the management range, logic (a sensing unit) for determining which assumed cause the deviation has occurred, an influence due to the deviation, a measure to be taken when the deviation has occurred, and an action for the measure, which are associated with each other and are enumerated inclusively. It should be noted that the knowledge base may be created not only by the HAZOP but also by FTA (Fault Tree Analysis), FMEA (Failure Mode and Effect Analysis), ETA (Event Tree Analysis), a method to which these are applied, a method similar to these, contents extracted from the results of listening to the operator, and contents extracted from the job standards and technical standards. In the present embodiment, abnormality sensing is performed based on parameters that are set to have causal relationships in the knowledge base.

Based on the information set in the table, the abnormal modulation cause specifying device 1 extracts data at a predetermined timing from the process data acquired from the plant unit 3, and performs an abnormality determination by a predetermined method. Fig. 10 is a diagram showing an example of a logical tree showing a relationship between modulation and its cause. The logical tree may be created based on the knowledge base shown in fig. 9. In the logic tree of fig. 10, the upstream and chronologically early events in the production process are arranged on the left side, and the downstream and chronologically subsequent events in the production process are arranged on the right side, and the layers are connected by arrows so as to go from the assumed cause to the modulation that appears as an influence. In the logical tree, when there are a plurality of assumed causes for one modulation in the knowledge base table, the assumed causes are connected in a branch manner, and events that appear in common from the assumed causes to the modulation are displayed in a bundle manner. The rectangles of the thick solid lines at the upstream end of each branch correspond to the assumed cause of the knowledge base table, and the numerals in parentheses in fig. 9 and 10 correspond to each other. In addition, the thin solid line rectangle corresponds to the influence of the knowledge base table, and represents an event that can be observed through the process data. For each of the influences, an operation is performed according to an operation method determined in the knowledge base table. Further, for each assumed cause, a model including a formula for performing the above-described operation may be defined, and the model is used to sense an abnormality or its precursor, or to assist in determination of the cause thereof.

< calculation method >

The operation described above may be, for example, an operation including the following method. Note that the abnormal modulation cause specification device 1 may display the calculation results of these methods.

Hotelling (T) ² Method)

For example, assuming that a plurality of process data obtained from one sensor obey a prescribed probability density function, the average and standard deviation of the population are estimated from the sample average and sample standard deviation calculated using the process data. The predetermined probability density function is, for example, a normal distribution. Then, the degree of abnormality is found based on the distance from the average value of the population to the process data of the verification object. The degree of abnormality is determined based on, for example, the square of the mahalanobis distance. The degree of abnormality based on the hotelling theory may be calculated using an instantaneous value of the process data itself, or using a maximum value, a minimum value, an integrated value, a standard deviation, a differential coefficient (slope), or the like of the process data for a predetermined period. According to the hotelling method, outliers from a specified reference can be sensed.

K nearest neighbor algorithm

For example vectorizing or matrixing time-series process data obtained from more than one sensor and calculating the distance of the data from each other. The distance may be an euclidean distance, a mahalanobis distance, or a manhattan distance. Then, the degree of abnormality is determined based on the distance from the data k-th closest to the data to be verified. In the k-nearest neighbor algorithm, the determination is made based on the relationship with other data. Therefore, for example, in a case where the normal value can be classified into a plurality of clusters, an outlier that is far from any of the plurality of clusters can be detected.

DTW (Dynamic Time warping) Barycenter Averaging

The averaged time series data is calculated based on a plurality of time series data such as process data in different batch processes. For example, the distance from the average time series data may be calculated for each process data of different production numbers in the corresponding section in the batch process. Fig. 11 is a diagram for explaining the synchronous processing of the process data. For each value of an element included in time series data of batch processes having different production numbers, the shortest distance between values included in different time series data is cyclically obtained, and the time series data is slid in the time axis direction to perform registration so that the integrated value of the shortest distances is minimized. That is, a plurality of time-series data are synchronized based on the similarity of the time-series data. In this way, a plurality of process data can be displayed in an overlapped manner so that steps in the processes performed in the unit 3 correspond in time series. Then, based on the integrated value of the distance between the synchronized time series data, the degree of abnormality is calculated by the k-nearest neighbor algorithm and the hotelling theory. According to DTW Barycenter Averaging, abnormalities can be sensed based on the degree of similarity between timing data.

The degree of abnormality of the plus or minus sign may be calculated for a deviation from a reference such as an average value. Fig. 12 is a diagram for explaining an example in which the degree of abnormality is calculated from the magnitude of the distance from the reference with respect to time-series data. Fig. 13 is a diagram for explaining an example of calculating the degree of abnormality from the distance from the reference in consideration of the positive and negative directions for the same time series data. In fig. 12 and 13, the vertical axis represents, for example, the degree of deviation from the average value. In the portion indicated by the rectangle of the dotted line, modulation is actually generated, but it is difficult to sense only from the values shown in the example of fig. 12. On the other hand, in the example of fig. 13, since the positive and negative directions are inclined to be opposite, the sensing of the modulation becomes easy.

For example, in the above-described hotelling method, the degree of deviation from the reference is obtained as a value to which a positive or negative sign is added without squaring the distance, and thereby the degree of abnormality as shown in fig. 13 is obtained. In DTW Barycenter Averaging, etc., a positive or negative sign is determined for a characteristic point such as a maximum value in time series data by the following equation, and the calculated value is multiplied by the magnitude of the distance.

Symbol decision = (μ -x)/| μ -x! non-conducting cells

μ is an average value (reference value) of the training data, and x is process data of the verification target. In this way, the sign indicating the direction of deviation from the reference at the time point can be determined according to the magnitude relationship between the reference value at the predetermined time point of the time series data and the process data to be verified at the corresponding time point, based on the sign determination equation. Further, by using a value with a sign indicating the degree of deviation from the reference, the degree of abnormality as shown in fig. 13 can be obtained, and erroneous sensing can be suppressed. In addition, as a characteristic point in the time series data, a maximum value, a minimum value, a difference between process data at a certain time point and process data at another time point, or the like may be used. Self-encoder (auto encoder)

Fig. 14 is a diagram for explaining abnormality sensing using a self-encoder. In the method, abnormality determination is performed based on characteristics of a relationship of process data from a plurality of sensors. Specifically, a model capable of compressing (encoding) and restoring (decoding) input data is created using a neural network by setting process data itself, which is, for example, continuous processing or batch processing of the input data, as a teaching value. In the neural network, for example, the number of nodes in the input layer and the output layer corresponds to the number of sensors, and the number of nodes in the intermediate layer is smaller than the number of sensors. Information input to the input layer is compressed in the intermediate layer and restored in the output layer. A plurality of intermediate layers may be present, and the connection structure between the layers is not limited to the full bonding. Then, a model is created in which parameters are adjusted so that the difference between the value of the input layer and the value of the output layer is small by performing learning processing using normal process data as training data. In the abnormality determination processing, process data to be verified is input, and an abnormality degree corresponding to a difference between the value of the input layer and the value of the output layer is calculated. That is, when process data having an abnormality is input, the information compressed in the intermediate layer cannot be appropriately restored in the output layer, and the difference between the values of the input layer and the output layer becomes large, so that abnormality sensing can be performed based on the difference. According to the self-encoder, an abnormality can be sensed based on the characteristic of the relationship of the output values among the plurality of sensors.

Graph lasso

For example, the dependency relationship between variables is expressed as a reference sparse graph by digitizing the dependency relationship based on a covariance matrix of process data from a plurality of sensors in a continuous process or a batch process. In a normal state, the dependency relationship between variables can be determined as a relationship that does not significantly deviate from the reference. Then, in the abnormality determination process, the dependency relationship between the variables is obtained using the process data to be verified, and the degree of abnormality corresponding to the magnitude of the difference from the above-described reference is calculated. According to the graph lasso, the correlation among the process data can be digitalized, and the degree of abnormality can be sensed based on the aliasing of the relationship.

In addition, a general abnormality sensing method or a method using the same may be used. The threshold value used for abnormality sensing in each method may be a value that is searched for as few as possible erroneous determinations in the normal case, that can quickly sense the occurrence of an abnormality in the abnormal case, and that is a sign thereof, using process data actually obtained during the operation of the unit 3, and may be registered in advance in a knowledge base shown in fig. 9.

< apparatus construction >

Fig. 15 is a block diagram showing an example of the configuration of the abnormal modulation cause specifying apparatus 1. The abnormal modulation cause specification device 1 is a general computer, and includes a communication interface (I/F) 11, a storage device 12, an input/output device 13, and a processor 14. The communication I/F11 may be, for example, a network card or a communication module, and communicates with another computer based on a predetermined protocol. The storage device 12 may be a main storage device such as a RAM (Random Access Memory) or a ROM (Read Only Memory) or an auxiliary storage device (secondary storage device) such as an HDD (Hard-Disk Drive), an SSD (Solid State Drive) or a flash Memory. The main storage temporarily stores a program read by the processor 14, information transmitted and received to and from another computer, or secures a work area of the processor 14. The auxiliary storage device stores a program executed by the processor 14, information transmitted and received to and from another computer, and the like. The input/output device 13 is, for example, a user interface such as an input device such as a keyboard or a mouse, an output device such as a monitor, or an input/output device such as a touch panel. The processor 14 is an arithmetic Processing device such as a CPU (Central Processing Unit), and executes a program to perform each process of the present embodiment. In the example of fig. 15, functional blocks are shown within processor 14. That is, the processor 14 functions as a process data acquisition unit 141, a preprocessing unit 142, a learning processing unit 143, an abnormality determination unit 144, a cause diagnosis unit 145, and an output control unit 146 by executing predetermined programs.

The process data acquisition unit 141 acquires process data from a sensor provided in the plant 3 via the communication I/F11 and the control station 2, for example, and stores the process data in the storage device 12. As described above, the process data is mapped to the sensor by the tag.

The preprocessing unit 142 processes the process data at the time of creating the abnormality sensing model. For example, the preprocessing section 142 associates the process data with the production number. That is, the process data corresponding to the predetermined label, system, and production number in the batch process and the process data corresponding to the predetermined label and output at the predetermined timing in the continuous process are associated based on the above-described trace back information held in the storage device 12 in advance. Further, data for a predetermined period for determining an abnormality is extracted based on a set value of a table such as a knowledge base, and a feature amount corresponding to each method is calculated. In the learning process, the preprocessing unit 142 may extract training data by performing data cleaning to exclude data during the non-steady operation period, data at the time of occurrence of an abnormality, and outliers such as noise.

The learning processing unit 143 creates an abnormality sensing model including one or more calculations based on, for example, a knowledge base, and stores the abnormality sensing model in the storage device 12. At this time, the learning processing unit 143 determines a parameter for learning the feature of the training data. Note that, in the case where the learning process is performed using the output values of a plurality of sensors, normalization may be appropriately performed.

The abnormality determiner 144 calculates the degree of abnormality using the process data and the abnormality sensing model.

That is, the abnormality determination unit 144 calculates the degree of abnormality using the test data for cross validation and the abnormality sensing model in the learning process. Further, in the abnormality determination process, the degree of abnormality is calculated using the process data acquired from the plant 3.

The cause diagnosing unit 145 calculates the degree of establishment (accuracy) for each of the plurality of assumed causes using the calculated degree of abnormality. The degree of satisfaction is calculated based on, for example, the proportion of the influence appearing in the process data among the influences associated with the respective assumed causes in the knowledge base, and the degree thereof, using the degree of abnormality calculated by the abnormality determination unit. In addition, it is also possible to present the action to the user by storing in advance the action indicating the countermeasure to be taken against the cause in the storage device 12 in association with each assumed cause.

The output control unit 146 issues an alarm or outputs the degree of satisfaction of each assumed cause when an abnormality is sensed via the input/output device 13, for example. The output control unit 146 appropriately connects the above-described components via the bus 15 in accordance with the user operation. For convenience, one of the apparatuses shown in fig. 15 includes the process data acquisition unit 141, the preprocessing unit 142, the learning processing unit 143, the abnormality determination unit 144, the cause diagnosis unit 145, and the output control unit 146, but at least some of the functions may be distributed among different apparatuses.

< learning Process >

Fig. 16 is a process flowchart showing an example of the learning process performed by the abnormal modulation cause specifying device 1. The processor 14 of the abnormal modulation cause specification apparatus 1 executes a predetermined program to execute processing as shown in fig. 16. The learning process is executed at an arbitrary timing using process data obtained from the past operation of the plant 3. The learning process mainly includes a preprocessing (fig. 16. That is, a part of the process data may be used as training data and the other part may be used as test data to perform cross-validation. The table or the like is a table created by the user and stored in the storage device 12 in advance. For convenience, although the preprocessing, the learning processing, and the verification processing are described in one processing flow shown in fig. 16, at least a part of the preprocessing, the verification processing, and the like may be distributed to different devices and executed.

The process data acquisition unit 141 of the abnormal modulation cause specification apparatus 1 acquires process data (fig. 16. In this step, data for the abnormality sensing model is extracted from the process data as shown in fig. 3 and 6. The Process data is data stored in the storage device 12 as OPC (Object Linking and Embedding for Process Control) data, a table called a database, or a file of a predetermined format such as CSV. The process data may include attributes such as date, time, and label, and may further include attributes such as a product number and a procedure particularly in the process data of the batch processing.

Further, the preprocessing unit 142 of the abnormal modulation cause determination device 1 associates the process data to be processed continuously with the production number (fig. 16. In this step, as shown in fig. 8, the process data acquired in the continuous process is associated with the production number group of the process data acquired in the batch process, and the process data used for the calculation of the degree of abnormality is associated with each other. That is, in the knowledge base shown in fig. 9 and the logic tree shown in fig. 11, when a certain cause affects both the process data of the batch process and the process data of the continuous process, the degree of abnormality and the degree of establishment are calculated based on the data related in this step.

Then, the preprocessing unit 142 extracts and processes data used in the abnormality determination model (fig. 16. In this step, the preprocessor 142 extracts data for a predetermined period for abnormality determination based on the set values of the table such as the knowledge base, and calculates the feature amounts according to the respective methods.

For example, when the degree of abnormality obtained by the hotelling method is calculated, the preprocessing unit 142 extracts process data at a predetermined timing or period, calculates an instantaneous value as the process data itself, a maximum value, a minimum value, an integrated value or difference of the process data, an integrated value of a reaction rate, a differential coefficient at a predetermined time point, and the like, and stores the calculated values in the storage device 12. In addition, when the degree of abnormality obtained by the k-nearest method is calculated, the time-series process data is vectorized or matrixed. In addition, when the abnormality degree obtained by DTW Barycenter Averaging is calculated, a plurality of process data are synchronized to obtain average time series data. In addition, when the degree of abnormality obtained by the self-encoder and the pattern lasso is calculated, a plurality of process data are synchronously processed.

The preprocessing unit 142 may perform predetermined data cleaning on the process data. The data cleaning process is a process of removing outliers, and various methods can be employed. For example, the moving average may be calculated using the most recent data. Further, the difference between the moving average and the actually measured value is taken, and the standard deviation σ indicating the variation of the difference is obtained. Then, for example, a value that does not fall within a predetermined reliable interval such as an interval from the average value of the probability distribution-3 σ to the average value of the probability distribution +3 σ (also referred to as a 3 σ interval) may be excluded. Similarly, values that do not fall within the 3 σ interval may be excluded from the difference between the actual measurement values before and after.

After that, the learning processing section 143 of the abnormal modulation cause determination device 1 performs the abnormal sensing model construction process (fig. 16. In this step, based on the knowledge base shown in fig. 9, an abnormality sensing model including an operation of the degree of abnormality is created. Specifically, the abnormality degrees obtained by the methods registered in the "arithmetic method" are calculated for one or more "influences" corresponding to the respective "assumed causes" in fig. 9, and an abnormality sensing model is created that shows a combination of the abnormality degrees. Further, the learning processing section 143 adjusts the parameters of the model by a method of abnormality sensing and using the training data. For example, when the degree of abnormality obtained by the encoder is calculated, the weighting coefficients between layers are adjusted so that the information of the process data to be input can be restored after compression. When the degree of abnormality obtained by graph lasso is calculated, the dependency relationship between variables is quantified based on the covariance matrix of process data from a plurality of sensors. Then, the learning processing section 143 causes the created abnormality sensing model to be stored in the storage device 12.

The abnormality determination section 144 of the abnormal modulation cause determination apparatus 1 performs calculation of the degree of abnormality using the created abnormality sensing model and the test data (fig. 16. In this step, the abnormality determination unit 144 calculates the degree of abnormality according to the method of calculating the degree of abnormality. For example, when calculating the degree of abnormality by the hotelling method, the process data is used to estimate the average value and standard deviation of the sample of the population, and the degree of abnormality is determined based on the distance from the average value of the population to the process data of the verification target. When the degree of abnormality obtained by the k-nearest neighbor algorithm is calculated, the distance between data is calculated, and the degree of abnormality corresponding to the distance to the data k-th nearest to the data to be verified is calculated. When the degree of abnormality obtained by DTW Barycenter Averaging is calculated, the degree of abnormality is obtained by the k-nearest neighbor algorithm or the hotelling theory based on the integrated value of the distance between time series data synchronized in the preprocessing. When the degree of abnormality obtained by the self-encoder is calculated, process data to be verified is input to the self-encoder, and the degree of abnormality corresponding to the difference between the value of the input layer and the value of the output layer is obtained. When the degree of abnormality obtained by graph lasso is calculated, the dependency relationship between variables is obtained using the process data to be verified, and the degree of abnormality corresponding to the magnitude of the difference in dependency relationship that becomes the reference is obtained.

The cause diagnosing unit 145 of the abnormal modulation cause specifying device 1 obtains the degree of establishment of the assumed cause using the calculated degree of abnormality (fig. 16. In this step, the degree of satisfaction is calculated based on the ratio of the occurrence of the corresponding modulation as an influence for the assumed cause of each knowledge base. For example, the reason (2) of fig. 9 corresponds to three influences of an increase in moisture of the tag 002, an increase in temperature 1 of the tag 004, and a decrease in temperature 2 of the tag 005. The degree of abnormality calculated for each influence in S31 in fig. 16 may be used to set the proportion of the influence in which the degree of abnormality exceeds the threshold value, among the three influences, as the degree of satisfaction. If the degree of abnormality of two of the three influences exceeds the threshold, the degree of satisfaction may be set to 66.7%, for example. In the calculation of the degree of satisfaction, weighting may be performed according to the type of influence (label) or the magnitude based on the degree of abnormality. For example, the satisfaction may be obtained by multiplying each influence by a weight and then summing the results.

Further, the output control unit 146 outputs the degree of abnormality calculated in S31 and the degree of satisfaction calculated in S32 (fig. 16. In this step, cross-validation is performed using test data different from training data used for model construction, among process data collected in the past operation of the plant 3. In this step, the process data at the time point when the abnormality occurred in the past is also used to appropriately sense the abnormality and verify whether or not an alarm and an action for coping with the abnormality are output. Further, the learning processing section 143 determines whether or not the abnormality is sensed with sufficient accuracy (fig. 16. If it is determined that the accuracy is insufficient (no in S4), the threshold value registered in the knowledge base (in other words, the normal range of the process data) is corrected so that the abnormality can be appropriately sensed, and the process from S31 onward is repeated. If it is determined in S4 that the abnormality can be sensed with sufficient accuracy (yes in S4), the operation using the abnormality sensing model and the threshold value created in S2 is performed. At least a part of the determination in S4 may be performed by the user.

The operation is, for example, an operation to be performed by the operator of the plant 3 in order to cope with the assumed reason is stored in the storage device 12 in advance in association with the assumed reason. Fig. 17 is a diagram showing an example of the action table. The table of fig. 17 includes attributes of reason, action 1, and action 2. In the column of the reason, a reason corresponding to an assumed reason of the knowledge base is registered. In the fields of action 1 and action 2, information indicating measures to be taken by the operator of the unit 3 to eliminate the corresponding cause is registered.

< abnormality sensing processing >

Fig. 18 is a process flowchart showing an example of the abnormality sensing process performed by the abnormality modulation cause determination device 1. The processor 14 of the abnormal modulation cause specification apparatus 1 executes a predetermined program to execute processing as shown in fig. 18. As for the abnormality sensing process, the process data obtained by the operation of the plant 3 is used to be executed substantially in real time. The abnormality sensing process mainly includes a preprocessing (fig. 18, s10), a model reading process (S20), and an abnormality determination process (S30). In fig. 18, the same reference numerals are given to steps corresponding to the learning process shown in fig. 16, and differences from the learning process will be mainly described below. For convenience, the description will be given as a process realized by the same apparatus as the apparatus that performs the learning process, but the apparatus that performs the abnormality sensing process may be an apparatus different from the apparatus that performs the learning process. Note that tables of the abnormality sensing model, the threshold value, the knowledge base, and the like created in the learning process are stored in the storage device 12 in advance.

The process data acquisition unit 141 of the abnormal modulation cause specification device 1 acquires the process data (fig. 18. The process data is data stored in the storage device 12 as OPC data, a table called a database, or a file having a predetermined format such as CSV. This step is substantially the same as S11 of fig. 16, but data relating to the process in operation is acquired in the unit 3. Further, the preprocessing section 142 of the abnormal modulation cause determination apparatus 1 associates the continuously processed process data with the production number (fig. 18. This step is the same as S12 in fig. 16. Then, the preprocessing unit 142 extracts and processes data used in the abnormality determination model (fig. 18. This step is substantially the same as S13 in fig. 16, but data cleaning is not necessary.

After that, the abnormality determination section 144 of the abnormal modulation cause specification device 1 reads out the abnormality sensing model created in the learning process from the storage device 12 (fig. 18. Further, the abnormality determination section 144 performs calculation of the degree of abnormality using the created abnormality sensing model and the process data obtained by the operation of the plant 3 (fig. 18. This step is the same as S31 in fig. 16. The cause diagnosing unit 145 of the abnormal modulation cause specifying device 1 obtains the degree of establishment of the assumed cause using the calculated degree of abnormality (fig. 18. This step is the same as S32 in fig. 16.

Further, the output control unit 146 outputs the degree of abnormality calculated in S31 and the degree of satisfaction calculated in S32, and issues an alarm when either degree of abnormality exceeds a predetermined threshold (fig. 18. In this step, the process data indicating the operation state of the plant 3, the degree of abnormality, and the degree of establishment of the assumed cause are presented to the user via the input/output device 13.

Fig. 19 is a diagram showing an example of a screen output to the input/output device 13. Fig. 19 is an example of a main control diagram, and shows transition of individual process data by a line graph. The area 131 displayed on the input/output device 13 displays a combination of the identification information and the latest value of a plurality of pieces of process data acquired from the plant 3. In the management map of the area 132, the transition of the value is shown by a line graph for the determined process data. The vertical axis represents the value of the process data, and the horizontal axis represents the time axis. In the example of fig. 19, the solid line represents the true value, and the broken line represents the estimated value. The true value is the process data itself to be subjected to the calculation of the degree of abnormality, and the estimated value may be an estimated value obtained by regression analysis of the process data to be subjected to the calculation of the degree of abnormality. The thin broken line is set as a line representing the upper and lower limits of the normal range (in other words, the threshold value for abnormality sensing). As indicated by circles in fig. 19, when the user operates the input/output device 13 such as a pointing device to move the pointer on the graph, the numerical value of the process data at the time point indicated by the pointer may be displayed. In the factor effect graph of the area 133, the horizontal axis shows the cause of modulation of the process data displayed in the area 132 or a label that can specify the cause, and the vertical axis shows the degree of establishment of the cause in a bar graph. This indicates that the larger the degree of establishment, the larger the possibility of modulation of the process data. The cause diagnosing unit 145 calculates the degree of satisfaction on the basis of the degree of abnormality calculated by the abnormality determining unit 144 for an event serving as an assumed cause of the modulation of the process data. The user can identify the cause of the modulation candidate and the accuracy thereof based on the magnitude of the degree of establishment, and can easily determine the modulation cause. The factor effect map is a map that the abnormality determination unit 144 calculates the degree of abnormality at the designated time or the current time and displays the degree of abnormality by the output control unit 146 when the "diagnosis" button in the region 134 is pressed.

Then, when the user operates the input/output device 13 such as a pointing device and selects any one of the bar charts of the factor effect graph, the cause of the modulation corresponding to the bar chart is displayed in the logic tree in a highlighted manner.

Fig. 20 is a diagram showing another example of a screen outputted to the input/output device 13 by the output control unit 146. Fig. 20 is an example of a tree diagram showing a logical tree like that shown in fig. 10. For example, in the case where the bar chart of the tag 004 is selected in fig. 19, the influence corresponding to the process data of the tag 004 is highlighted on the logic tree. The highlight display is completed by changing the display form, for example, changing the color or the line type. In fig. 20, shading is applied to the corresponding rectangle. Further, a rectangle of a thick line connected to the upstream side of the logical tree represents an assumed cause of influence. The causes may be displayed as outlined in fig. 20, or the effects on the process data other than the causes may be displayed. Note that the satisfaction degrees of the causes calculated in S32 in fig. 18 may be displayed, or the operation may be displayed. Further, the cause may be displayed when the user moves the pointer on each rectangle.

When the "trend" button shown in fig. 19 and 20 is pressed, the process trend of each label shown in the factor effect diagram may be displayed, or the process trend of a label that can specify the cause of modulation may be particularly displayed. The process trend is a content that calculates a value for each period, for example, for each predetermined time, each predetermined number of days, each predetermined number of months, or each season, using the process data stored in the storage device 12, and indicates the value on the graph.

The output control unit 146 may output a log of the degree of abnormality at a timing when the degree of abnormality calculated by each calculation method exceeds a predetermined threshold value, for example. Note that a log of the assumed cause and the establishment degree may be output. Each log is associated with date and time, a production number, an arithmetic method, an abnormality sensing model, and the like and is output, thereby facilitating analysis of abnormal modulation.

< modification example >

The respective configurations and combinations thereof in the respective embodiments are examples, and addition, omission, replacement, and other modifications of the configurations may be appropriately made within the scope not departing from the gist of the present invention. The present disclosure is not limited by the embodiments, but only by the claims. In addition, the various aspects disclosed herein may also be combined with any of the other features disclosed herein.

In the above-described embodiments, the chemical unit is described as an example, but the present invention can be applied to a manufacturing process in general production equipment. For example, instead of the production number of the batch process in the embodiment, the batch number may be used as a processing unit and processing according to the batch process in the embodiment may be applied.

At least a part of the functions of the abnormal modulation cause specification apparatus 1 may be realized so as to be distributed among a plurality of apparatuses, or a plurality of apparatuses may provide the same functions in parallel. Note that at least a part of the functions of the abnormal modulation factor specifying device 1 may be provided on a so-called cloud (cloud).

Further, the present disclosure includes a method of executing the above-described processing, a computer program, and a computer-readable recording medium in which the program is recorded. The recording medium in which the program is recorded can perform the above-described processing by causing a computer to execute the program.

Here, the computer-readable recording medium refers to a recording medium that stores information such as data and programs by an electric, magnetic, optical, mechanical, or chemical action and reads the information from a computer. Among such recording media, those removable from the computer include floppy disks, magneto-optical disks, magnetic tapes, and memory cards. Further, as a recording medium fixed to a computer, there are an HDD, an SSD (Solid State Drive), a ROM, and the like.

Description of the reference numerals

1: abnormal modulation cause determining means;

11: a communication I/F;

12: a storage device;

13: an input/output device;

14: a processor;

141: a process data acquisition section;

142: a pretreatment section;

143: a learning processing unit;

144: an abnormality determination unit;

145: a cause diagnosis unit;

146: an output control section;

2: a control station;

3: and (4) a machine set.

Claims

1. An abnormal modulation cause display device includes:

a process data acquisition unit that reads out process data continuously output from a plurality of sensors provided in a production facility from a storage device that stores the process data;

an abnormality determination unit that calculates an abnormality degree indicating a degree of modulation of the process data read by the process data acquisition unit;

a cause diagnosing unit configured to determine whether or not the degree of abnormality calculated by the abnormality determining unit satisfies a predetermined criterion with respect to the process data output from the plurality of sensors, using cause-and-effect relationship information defining a combination of a cause and a modulation of the process data output from the plurality of sensors that appears as an influence of the cause; and

and an output control unit that reads out information indicating a countermeasure to be taken for the reason from the storage device that also stores the information indicating the countermeasure, and outputs the information to an output device.

2. The abnormal modulation cause display device according to claim 1,

the cause diagnosing unit obtains the accuracy of the cause of the modulation by multiplying the process data of the plurality of types by a coefficient corresponding to the type of the process data or a coefficient based on the magnitude of the abnormality degree,

the abnormal modulation cause display device further includes:

and an output control unit that outputs, to an output device, a plurality of candidate causes that can cause the modulation, and the accuracy of the causes, for the modulation.

3. The abnormal modulation cause display device according to claim 1 or 2,

the output control unit outputs, to the output device, a logical tree in which events occurring from the cause to the modulation are connected into layers, the logical tree having the modulation as a root and the cause of the modulation as a leaf, based on the causal relationship information, and outputs, to the output device, information indicating a measure to be taken for the cause in association with the cause.

4. A method for displaying abnormal modulation reason is executed by a computer as follows:

reading out the process data from a storage device that stores the process data continuously output from a plurality of sensors provided in the production facility,

an abnormality degree indicating a degree of modulation of the read process data is calculated,

determining whether or not the calculated abnormality degree satisfies a predetermined criterion with respect to the process data output from the plurality of sensors, using cause and effect information defining a combination of a cause and a modulation of the process data output from the plurality of sensors that appears as an influence due to the cause,

the information indicating the countermeasure to be taken for the reason is read from the storage device that also stores the information indicating the countermeasure and output to the output device.

5. An abnormal modulation cause display program for causing a computer to execute:

determining whether or not the calculated abnormality degree satisfies a predetermined criterion with respect to the process data output from the plurality of sensors, using causal relationship information defining a combination of a cause and a modulation of the process data output from the plurality of sensors that appears as an influence due to the cause,